Distinction advises and delivers digital work, so our view of agency relationships comes with an obvious commercial interest. The useful test is whether the structure below protects the client when expectations diverge, delivery becomes difficult or the relationship ends.

A strong portfolio and persuasive pitch do not make an engagement self-governing. Trouble usually develops in the rules around the work: ambiguous scope, absent decision rights, incentives that conflict with uncertainty, inaccessible project assets or a gap between the promised and assigned team.

Trust matters. Trust without structure is optimism, and optimism is a poor project control.

Test the working relationship before buying it

Meet the people expected to lead and deliver the work. Ask what proportion of their time is committed, which skills are assigned and what happens if a named person becomes unavailable. Allocation may legitimately remain provisional before contract, but the uncertainty should be visible.

Look beyond selected case studies. Speak to references about a project that encountered difficulty. Ask when the agency raised the issue, how options were presented, whether forecasts remained credible and how handover worked. A good result from a difficult engagement can reveal more than a polished launch.

During procurement, create a respectful disagreement. Ask the agency to explain an assumption, challenge a recommendation or compare a cheaper option. The aim is not performance theatre. It is to see whether specialists use evidence, acknowledge uncertainty and still respect the client’s authority.

Inspect the delivery system itself: a representative plan, decision record, quality approach, project board, risk process and example of forecast-to-complete. Methodology labels are less revealing than the artefacts used on an ordinary week.

Define the problem and boundaries together

Scope documents fail when both parties recognise a word and attach different meanings to it. “Content migration”, “integration”, “testing” and “launch support” can each conceal several kinds of work.

For every significant deliverable, define:

  • included activity and output;
  • client inputs and assumptions;
  • acceptance evidence and decision owner;
  • dependencies and exclusions;
  • quality, accessibility, security and performance expectations;
  • effect on time and cost if an assumption changes.

Record the outcome the engagement is meant to improve. Requirements will evolve as the team learns, so the relationship needs a controlled way to reconsider scope without treating all change as failure.

Change control should present options and consequences. A new requirement may replace existing scope, extend time, increase cost, reduce another commitment or wait for a later phase. Approving change without a trade-off makes the forecast fiction.

Match the commercial model to uncertainty

No pricing model guarantees aligned behaviour. Fixed price can suit well-understood work and encourage narrow interpretation when discovery remains substantial. Time and materials can support learning while placing more cost risk on the client. Retainers can provide continuity and obscure output if priorities and capacity are unclear.

Choose the model deliberately. Define rates, assumptions, expenses, invoicing, acceptance, forecast updates, change, cancellation and transition. Link milestone payments to evidenced outcomes or deliverables where appropriate, while avoiding a structure that rewards premature acceptance or withholds so much value that collaboration becomes adversarial.

Share the likely total horizon even when contracting in phases. A small first commitment is useful if it produces a coherent result and a real decision about what follows. It should not disguise an unaffordable later dependency.

Protect ownership and continuity

Contracts need specialist legal review. Operationally, the client should understand who owns or may use:

  • code, configuration and reusable components;
  • designs, research and content;
  • data and derived information;
  • accounts, domains and environments;
  • third-party licences and dependencies;
  • documentation and deployment automation.

Provide client-authorised access throughout delivery, subject to appropriate security controls. A project board, repository or design file should not appear only at handover. Confirm that another authorised person can build, deploy, operate and recover the service using current assets.

Define termination, notice, work in progress, data return and deletion, access removal, supplier cooperation and transition rates before they are needed. Clear exit terms protect a healthy relationship because neither party must avoid necessary decisions for fear of being trapped.

Establish two governance rhythms

Delivery governance handles near-term work: progress against the plan, forecast, blockers, quality evidence, decisions and risks. Its cadence should match the work rather than follow a universal weekly rule.

Steering governance examines direction: whether the original outcome remains valid, which trade-offs require sponsor authority, whether risk is acceptable and what the business has changed. Keep it distinct from routine task management.

Write decision rights in plain language. A RACI can help, but a list of named decisions and accountable owners is often clearer. Specify response times for client inputs because late approval or missing content affects delivery even when the cause sits outside the agency.

Escalation should be predictable and proportionate. The delivery leads try to resolve an issue with evidence and options. Matters beyond their authority move to named sponsors. Safety, legal or ethical concerns follow the appropriate specialist route immediately.

Govern without directing every task

The client owns outcomes, risk acceptance and material decisions. The agency owns the professional quality of its advice and delivery within the agreement. Micromanagement confuses those roles and can hide whether the supplier is actually capable.

Ask for evidence rather than constant permission gates. Review working software, research, test results, forecasts and decision logs. Attend sessions where client knowledge is essential. Allow specialists to choose implementation details within agreed architecture and controls, with proportional independent assurance where consequence warrants it.

Good collaboration is usually “done with you” for services the client must own and evolve. Client colleagues contribute knowledge and decisions while building operational capability. A more hands-off model may suit a contained commodity output. Choose based on ownership after launch, risk and internal capacity rather than agency positioning.

Read patterns, not isolated bumps

Every project encounters difficulty. Material warning patterns include:

  • forecast changes arriving after the decision window;
  • repeated unexplained team substitutions;
  • additional scope presented without consequence or options;
  • long gaps in working evidence;
  • serious defects ageing without an owner;
  • senior oversight promised and then absent;
  • persistent disagreement about who delayed what;
  • resistance to authorised access or transition preparation.

Respond by checking records and asking for a recovery plan. Separate client dependencies from supplier performance. Increase assurance, change scope, alter access or pause work according to consequence. Do not wait until the only remaining option is a crisis meeting.

The client also has obligations. Provide decisions, people, access, content and feedback when agreed. Surface internal disagreement early. An agency relationship cannot compensate indefinitely for missing business ownership.

Before signing, test the named team, scope boundaries, commercial assumptions, decision rights, quality evidence, asset access and exit path. During delivery, keep forecasts and working evidence visible. At transition, verify that the organisation can operate what it has bought.

If you are preparing a significant digital programme, Distinction’s assessment can help clarify the problem and engagement model before a build commitment. The same questions are worth asking whether the eventual partner is Distinction or another firm. More information is available at distinction.co.uk.